平台
Reliability hardening
Error monitoring, queue resilience, and graceful degradation across the platform.
What shipped. Error monitoring across the whole platform, retry-safe background queues for AI generation and publishing, and graceful degradation so a failing subsystem dims a feature instead of taking a page down.
Why it matters. A publishing platform earns trust by being boring: chapters load, uploads land, generation jobs finish or tell you why not.
How to use it. You should notice nothing — that is the feature. If something does go wrong mid-generation, the job reports its state instead of silently disappearing.